Grad Student Solves Astronomical Optics Problems Using AI

 

Credit: Robin Swanson, Dunlap Institute.

A University of Toronto graduate student is using artificial intelligence to solve one of the biggest technical challenges in astronomy.

Robin Swanson – a PhD student in U of T’s Department of Computer Science and the Dunlap Institute for Astronomy & Astrophysics – is applying artificial intelligence to astronomical instrumentation in order to correct for the twinkling of stars in the Earth’s atmosphere.

The problem he has helped to solve, to put simply, is actually caused by the air. “There is a lot of air between the telescope and whatever distant star or galaxy we are trying to image,” Swanson explains. “That wouldn’t be an issue, except different parts of the air have different temperatures that bend the light in an unwanted way. This causes the image to be blurry.”

To solve this, Swanson and his team have trained an image processing algorithm to take in a history of previous atmospheric conditions, and then use that history to help predict what those conditions will look like in the future.

“Basically we can apply those predictions to our adaptive optics to undo the effects of the atmosphere.”

Swanson assembling the adaptive optics testbed with another student. Credit: Masen Lamb, Dunlap Institute.

The results of this research have been published in the Monthly Notices of the Royal Astronomical Society (MNRAS). Swanson is the lead author on the publication, along with Dunlap members Dr. Masen Lamb and Professor Suresh Sivanandam.

Sivanandam says Swanson found unique solutions to make this research possible. “We knew going in that the problem was amenable to artificial intelligence, because of existing research,” Sivanandam says.

“Robin’s approach uses the latest innovations in machine learning algorithms – and he has a clear path for implementation on existing astronomical telescopes.”

Sivanandam calls Swanson “the heart and soul” of the research.

The team is already working hard toward next steps. While at this point their results have come from simulations, they’re now gearing up to run their algorithm on real hardware.

In the longer term, Swanson says the possibilities are vast. “This new capability could allow for anything from helping to find new exoplanets, to imaging new regions of space.”

 

Associated Publication Info: 

Swanson, R.; Lamb, M.; Correia, C.M.; Sivanandam, S.; Kutulakos, K. Closed loop predictive control of adaptive optics systems with convolutional neural networks. Monthly Notices of the Royal Astronomical Society, Volume 503, Issue 2, May 2021, Pages 2944-2954. https://ui.adsabs.harvard.edu/abs/2020ApJ…901..135O/abstract

 

For more information, please contact:
Meaghan MacSween
Communications and Multimedia Officer
Dunlap Institute for Astronomy & Astrophysics,
University of Toronto
meaghan.macsween@utoronto.ca

 

The Dunlap Institute for Astronomy & Astrophysics at the University of Toronto is an endowed research institute with more than 90 faculty, postdocs, students and staff, dedicated to innovative technology, ground-breaking research, world-class training, and public engagement. The research themes of its faculty and Dunlap Fellows span the Universe and include: optical, infrared and radio instrumentation; Dark Energy; large-scale structure; the Cosmic Microwave Background; the interstellar medium; galaxy evolution; cosmic magnetism; and time-domain science. The Dunlap Institute for Astronomy and Astrophysics, David A. Department of Astronomy & Astrophysics and the Canadian Institute for Theoretical Astrophysics comprise the leading centre for astronomical research in Canada, at the leading research university in the country, the University of Toronto.